Observer agreement in assessing flexion-extension X-rays of the cervical spine, with and without the use of quantitative measurements of intervertebral motion

Abstract

Background context Flexion-extension X-rays are commonly used to identify abnormalities in intervertebral motion, despite little evidence for the reliability of the information that clinicians derive from these test. Purpose Quantify observer agreement on intervertebral motion abnormalities assessed with and without the use of computer-assisted technology. Study design Assess interobserver agreement among clinicians when they evaluate cervical flexion-extension X-rays using the methods they now use in clinical practice, and compare this to observer agreement when the same clinicians reassess the X-rays using computer-assisted technology. Methods Seventy-five flexion-extension X-rays of the cervical spine, obtained from several clinical practices, were assessed by seven practicing physicians who routinely assess these X-rays. Observers assessed the studies using the methods they routinely use, and then reassessed the studies, at least a month later, using validated computer-assisted methods. Agreement among clinicians with and without computer-assisted technology was assessed using kappa statistics. Results Agreement was poor (kappa=0.17) with methods routinely used in clinical practice. Computer-assisted analysis improved interobserver agreement (kappa=0.77). With computer-assisted methods, disagreements involved cases with severe degeneration or static misalignment where motion was within normal limits, or in fusion cases where there was between 1 and 1.5 degrees of motion at the fusion site. Conclusions This study suggests that commonly used methods to assess flexion-extension X-rays of the cervical spine may not provide reliable clinical information about intervertebral motion abnormalities, and that validated, computer-assisted methods can dramatically improve agreement among clinicians. The lack of definitions of instability and fusion acceptable to all the clinicians was likely a primary source of disagreement with both manual and computer-assisted assessments.